市场前瞻 · Market Outlook

纵观有记载的历史，任何时期的头条新闻都呈现出同样的模式：当时最重要的公司主导着当时的讨论。17世纪，荷兰东印度公司控制着贸易路线、货币和整个政府。19世纪末，标准石油公司和铁路公司——卡内基、洛克菲勒、范德比尔特——引发了与今天特斯拉和马斯克一样的愤怒、同样的痴迷和同样的铺天盖地的报道。上世纪之交，美国钢铁公司是历史上第一家市值十亿美元的公司，其统治地位似乎不可撼动。大萧条时期，IBM是少数几家在其他公司纷纷倒闭时仍能保持增长的公司之一。20世纪50年代，通用汽车成为当时全球规模最大、盈利能力最强的公司。如今，特斯拉、英伟达以及其他正在打造下一代产品的公司引领潮流。只有名字变了。

Throughout recorded history, the same pattern repeats: the most important companies of any era dominate the conversation of that era. In the 17th century, the Dutch East India Company controlled trade routes, currencies, and entire governments. In the late 19th century, Standard Oil and the railroads — Carnegie, Rockefeller, Vanderbilt — generated the same fury, the same obsession, and the same wall-to-wall coverage as Tesla and Musk do today. At the turn of the last century, US Steel was the first company in history to reach a billion-dollar market cap, its dominance seemingly unassailable. During the Great Depression, IBM was one of the few companies that continued to grow while others collapsed. In the 1950s, General Motors became the largest and most profitable company in the world. Today, Tesla, Nvidia, and others building the next generation of products are leading the way. Only the names change.

Waymo与Tesla：一场关于数据的竞争

Waymo vs Tesla: A Competition About Data

团队留意到，不同的投资者对Waymo与Tesla这两家公司往往持有极端对立的看法。作为投资分析团队，职责不在于偏袒任何一方，而在于分析数据，帮助会员在复杂的市场环境中作出更有依据的判断。以下是团队基于现有数据的客观梳理。

The team notes that investors tend to hold sharply opposing views on Waymo and Tesla. As an investment analysis team, the role is not to take sides, but to analyse the data and help members make better-informed decisions in a complex market environment. The following is the team's objective assessment based on available data.

将Waymo与Tesla FSD直接比较，实际上是在对比两种截然不同的策略。Waymo目前在地理围栏城市内运营数百辆无人驾驶出租车，而Tesla则拥有数百万辆汽车每天在全球各地实时收集真实驾驶数据。两者并非同一回事。

Comparing Waymo directly to Tesla FSD is in effect comparing two fundamentally different strategies. Waymo currently operates several hundred driverless taxis within geofenced cities, while Tesla has millions of vehicles collecting real-world driving data in real time across the globe every day. These are not the same thing.

数据层面，Tesla FSD车队已累计行驶约85亿英里，平均每天新增约2000万英里，预计本月中旬突破100亿英里。相比之下，Waymo累计行驶里程约为2亿英里，仅占Tesla总里程的约2.2%。Tesla累计行驶里程约为Waymo的45倍。Waymo只能在少数几个预先绘制好地图、经过精心协商获得许可的受控区域内运营，而Tesla则在世界各地、各种路况下行驶，实时训练系统。这并非势均力敌的竞争，而是两种完全不同规模与策略的布局。

On the data front, Tesla's FSD fleet has accumulated approximately 8.5 billion miles, adding roughly 20 million miles per day, with the 10 billion mile milestone expected mid-month. By comparison, Waymo's cumulative mileage stands at approximately 200 million miles — roughly 2.2% of Tesla's total. Tesla's accumulated mileage is approximately 45 times that of Waymo. Waymo operates only within a handful of carefully mapped and permitted controlled zones, while Tesla drives across all road conditions worldwide, training the system in real time. This is not an evenly matched competition — it is two entirely different scales and strategies playing out simultaneously.

Tesla目前未能实现规模化运营的核心障碍，并非技术本身，而是法律责任、监管体系与游说阻力。当监管许可与现有技术相匹配时，Tesla的部署规模将与Waymo的出租车示范完全不同。这是一种截然不同的策略，规模亦截然不同。

The primary obstacle to Tesla achieving scaled operations today is not the technology itself, but legal liability, regulatory frameworks, and lobbying resistance. When regulatory approvals align with existing technology, Tesla's deployment scale will be fundamentally different from Waymo's taxi demonstrations. This is a completely different strategy, operating at a completely different scale.

关于Grok与AI模型竞争

On Grok and the AI Model Competition

AI并非一场单一的竞赛，而更像是十场不同的比赛同时进行。在AIME 2025数学评估中，Grok 3的准确率达93.3%；在GPQA科学评估中准确率达84.6%；在情商基准测试中，Grok 4.1以1586分位居榜首。这些数字并不像一个"落后得离谱"的平台该有的表现。

AI is not a single race — it is more like ten different races happening simultaneously. In the AIME 2025 mathematics evaluation, Grok 3 achieved an accuracy rate of 93.3%. In the GPQA science evaluation, it reached 84.6%. In emotional intelligence benchmarks, Grok 4.1 ranked first with a score of 1,586. These are not the numbers of a platform that is hopelessly behind.

Grok最擅长实时市场研究与社交媒体情报分析，能够从X数据中提取情感分析、趋势识别与人口统计洞察，这是其他模型在不拥有数据流的情况下无法复制的结构优势。不同的AI模型在不同的任务上各有所长，关键在于理解每个工具的适用场景，而非简单地以单一标准评判优劣。

Grok excels at real-time market research and social media intelligence, extracting sentiment analysis, trend identification, and demographic insights from X data — a structural advantage that competing models cannot replicate without owning the data stream. Different AI models have different strengths across different tasks. The key is understanding the appropriate use case for each tool, rather than applying a single standard to judge them all.

关于马斯克的争议性与Tesla的价值创造

On Musk's Controversy and Tesla's Value Creation

团队承认，关键人物风险是真实存在的，马斯克的争议性亦无可否认。然而，投资决策的核心依据是业绩记录而非个人喜好。以下是数据所呈现的事实：Tesla在其鼎盛时期每年为投资者创造的价值约为790亿美元。贝佐斯用26年将亚马逊打造成万亿美元公司，乔布斯约14年达到约3000亿美元市值巅峰，而马斯克在同一时间内创造的市值是乔布斯的三倍以上。在有记载的历史中，没有人能以同样的速度与规模完成这样的价值创造。福特曾引起争议，乔布斯也曾引起争议，贝佐斯至今依然如此。市场不会因为你受欢迎就付钱，而是会因为你的业绩就付钱。

The team acknowledges that key person risk is real, and Musk's controversial nature is undeniable. However, investment decisions are grounded in performance records, not personal preference. Here is what the data shows: Tesla created approximately $79 billion in investor value per year at its peak. Bezos took 26 years to build Amazon into a trillion-dollar company; Jobs reached a peak market cap of approximately $300 billion in around 14 years; Musk created more than three times Jobs' market cap in the same timeframe. In recorded history, no one has combined speed and scale of value creation in this way. Ford was controversial, Jobs was controversial, and Bezos remains so today. The market does not pay for popularity — it pays for performance.

投资者不必认同马斯克的个人行为或公开立场，这是个人选择。团队的职责是分析数据并呈现客观事实，最终决定权始终在每位会员自己手上。

Investors do not need to agree with Musk's personal conduct or public positions — that is a personal choice. The team's role is to analyse data and present objective facts. The final decision always rests with each individual member.

微软（MSFT）：与OpenAI关系重组，核心逻辑不变

Microsoft (MSFT): OpenAI Relationship Restructured, Core Thesis Unchanged

市场对微软与OpenAI"分手"的描述存在重大误读。正确的理解是：微软将停止向OpenAI支付收入分成，同时保留OpenAI知识产权的免版税许可直至2032年。换言之，微软获得了资产，同时省去了成本。

The market's characterisation of the Microsoft-OpenAI "split" contains significant misreading. The accurate understanding is this: Microsoft will cease paying OpenAI a revenue share while retaining a royalty-free licence to OpenAI's intellectual property through 2032. In other words, Microsoft acquired the assets while eliminating the cost.

最新季度数据显示，Azure本季度增长40%，AI年度营收已超370亿美元，同比增长123%。微软目前拥有超过2000万个付费Copilot席位，同比增长250%。Azure AI Foundry现包含来自Cohere、DeepSeek、Meta、Mistral、xAI等合作伙伴的逾11000个模型，财富500强企业中已有80%在使用该平台。此外，根据新协议，OpenAI已签约购买价值2500亿美元的Azure服务，这意味着无论独家协议如何变更，微软未来多年的收入均已得到保障。

The latest quarterly data shows Azure grew 40% this quarter, with AI annualised revenue surpassing $37 billion, up 123% year-over-year. Microsoft now has over 20 million paid Copilot seats, up 250% year-over-year. Azure AI Foundry now includes over 11,000 models from partners including Cohere, DeepSeek, Meta, Mistral, and xAI, with 80% of Fortune 500 companies already using the platform. Furthermore, under the new agreement, OpenAI has committed to purchasing $250 billion worth of Azure services, meaning Microsoft's revenue for many years ahead is secured regardless of changes to the exclusivity arrangement.